The endogenous siRNA pathway is involved in heterochromatin formation in Drosophila ´ Delphine Fagegaltiera,1,2, Anne-Laure Bouge´a,1, Bassam Berrya, Emilie Poisotb, Odile Sismeiroc, Jean-Yves Coppe´ec, Laurent The´odoreb, Olivier Voinnetd, and Christophe Antoniewskia,3 aCentre

National de la Recherche Scientifique, Unite´ de Recherche Associe´e 2578, Institut Pasteur, 25 rue du Dr Roux, F75015 Paris, France; bCentre National de la Recherche Scientifique, Unite´ Mixte de Recherche, E´cole Pratique des Hautes E´tudes 8159, Universite´ Versailles Saint Quentin, 45 avenue des Etats-Unis, F78035 Versailles, France; cGe´nopole, Institut Pasteur, Plate-forme 2, 28 rue du Dr Roux, F-75015 Paris, France; and dCentre National de la Recherche Scientifique, Unite´ Propre de Recherche 2357, Institut de Biologie Mole´culaire des Plantes, 12 Rue du Ge´ne´ral Zimmer, F67084 Strasbourg Cedex, France Communicated by Jules A. Hoffmann, Centre National de la Recherche Scientifique, Strasbourg, France, September 15, 2008 (received for review July 3, 2008)

A new class of small RNAs (endo-siRNAs) produced from endogenous double-stranded RNA (dsRNA) precursors was recently shown to mediate transposable element (TE) silencing in the Drosophila soma. These endo-siRNAs might play a role in heterochromatin formation, as has been shown in S. pombe for siRNAs derived from repetitive sequences in chromosome pericentromeres. To address this possibility, we used the viral suppressors of RNA silencing B2 and P19. These proteins normally counteract the RNAi host defense by blocking the biogenesis or activity of virus-derived siRNAs. We hypothesized that both proteins would similarly block endo-siRNA processing or function, thereby revealing the contribution of endo-siRNA to heterochromatin formation. Accordingly, P19 as well as a nuclear form of P19 expressed in Drosophila somatic cells were found to sequester TE-derived siRNAs whereas B2 predominantly bound their longer precursors. Strikingly, B2 or the nuclear form of P19, but not P19, suppressed silencing of heterochromatin gene markers in adult flies, and altered histone H3-K9 methylation as well as chromosomal distribution of histone methyl transferase Su(var)3–9 and Heterochromatin Protein 1 in larvae. Similar effects were observed in dcr2, r2d2, and ago2 mutants. Our findings provide evidence that a nuclear pool of TE-derived endo-siRNAs is involved in heterochromatin formation in somatic tissues in Drosophila. RNAi 兩 nucleus 兩 viruses

R

ecent deep sequencing efforts have provided critical information on Drosophila small RNA repertoires in various tissues and during distinct developmental stages (1–7). Four classes of small RNAs mediate posttranscriptional gene silencing in Drosophila: i) ⬇22-nt miRNAs are processed from stem-loop precursors by Dicer-1 and repress mRNA expression; ii) ⬇25-nt piRNAs are produced from transposable element (TE) transcripts in gonads where they silence TEs through a feedback regulatory mechanism involving the PIWI subfamily of Argonautes (2, 3, 8–11); iii) 21-nt siRNAs are processed from long dsRNAs by Dicer-2 and trigger RNAi, for instance in response to viral infection (12–14); and iv) recently discovered 21-nt endo-siRNAs are processed from endogenous dsRNA precursors by Dicer-2 and silence TEs, and possibly endogenous mRNA in somatic tissues (1, 5–7, 15). In S. pombe, siRNAs produced from repetitive sequences in chromosome pericentromeres direct heterochromatin formation and transcriptional gene silencing. As in S. pombe (16), Drosophila heterochromatin is prominent in pericentromeric regions, mostly comprised of short satellite repeats and TEs, and is associated with histone H3 methylation on lysine 9 (H3K9) by the histone methylase Su(var)3–9 (Clr4 in S pombe). This allows recruitment of the Heterochromatin Protein HP1 (SWI6 in S. pombe) to maintain and spread heterochromatin to nearby genes (17). Despite these analogies, the evidence supporting a role of small RNAs in heterochromatin formation and transcriptional gene silencing in Drosophila remain indirect (18, 19). Mutants for the Argonautes Piwi and Aubergine or for the RNA helicase

21258 –21263 兩 PNAS 兩 December 15, 2009 兩 vol. 106 兩 no. 50

Spindle-E exhibit decreased H3-K9 methylation, altered recruitment of HP1 and decreased silencing of heterochromatin markers and of several classes of TEs (20–22); Piwi was shown to interact directly with HP1 (23). In addition, it is noteworthy that these data point out piRNAs that are mostly produced in gonads, suggesting that this class of small RNAs play an initiator role in heterochromatin establishment in the germ line. Here, we show that another class of siRNA derived from TE transcripts, endo-siRNAs, plays a role in heterochromatin formation in somatic tissues during larval development and in adults. Our data strongly suggest that proper nuclear localization of these siRNAs is essential to regulate chromatin dynamics in Drosophila. Results and Discussion We examined the length distribution of TE-matching small RNAs in publicly available small RNA libraries from the fly soma (4), see Materials and Methods). We found that a dramatic shift in the size of repeat-derived small RNAs occurs during development: the greater population of ⬇25 nt species detected in very early embryos, largely composed of maternally deposited TE-derived piRNAs (24), is replaced by a population of ⬇21 nt species in pupae, adult heads and S2 cells (Fig. 1). This size shift is consistent with previous observations indicating that TE-derived siRNAs are produced in somatic tissues (1). Whether these endogenous TE-derived siRNAs are involved in heterochromatin formation in the soma, however, remains unanswered (25). To address this question, viral proteins known to counteract antiviral RNAi were expressed in flies and their effects on endogenous TE-derived siRNAs were assessed in the soma. The Tombusvirus P19 and Flock House virus B2 proteins suppress antiviral RNAi in plants and insects, respectively (26). P19 forms a head-to-tail homodimer that specifically sequesters siRNA duplexes (27–29), whereas B2 forms a four-helix bundle that binds to one face of an A-form RNA duplex, independent of its length. As a consequence, and unlike P19, B2 prevents the processing of long dsRNAs into siRNAs by the Drosophila Dicer-2 (30–33). We found that silencing of endogenous white or EcR genes by invertedrepeat constructs (34, 35) is suppressed in transgenic adults expressing B2 or P19 in the eye (Fig. S1 A–B). In contrast, P19 fused to a nuclear localization peptide (NLS-P19) (Fig. S2 A) barely Author contributions: D.F., A.-L.B., and C.A. designed research; D.F., A.-L.B., B.B., E´.P., and O.S., performed research; D.F., A.-L.B., B.B., and O.V. contributed new reagents/analytical tools; D.F., A.-L.B., B.B., E´.P., O.S., J.-Y.C., L.T., O.V., and C.A. analyzed data; and D.F., A.-L.B., O.V., and C.A. wrote the paper. 1D.F.

and A.-L. B. contributed equally to this work.

2To

whom correspondence may be addressed at : Cold Spring Harbor Laboratory, 1 Bungtown Road, Cold Spring Harbor, NY 11724. E-mail: [email protected].

3To

whom correspondence may be addressed at: Drosophila Genetics and Epigenetics, Institut Pasteur, 25 rue du Dr Roux, F75015 Paris, France. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/cgi/content/full/ 0809208106/DCSupplemental.

www.pnas.org兾cgi兾doi兾10.1073兾pnas.0809208105

Very early embryos

150 100

Number of reads in a thousand

50 0 40 35 30 25 20 15 10 5 0 8 7 6 5 4 3 2 1 0

20

Late embryos

Adult heads

15 10 5 0 200

Pupae

S2 cells

150 100 50 0 19 20 21 22 23 24 25 26 27 28 29 19 20 21 22 23 24 25 26 27 28 29 small RNA length (nt)

Fig. 1. The length of TE-matching small RNAs varies during Drosophila development. Length distribution of the TE-matching small RNAs sequences from very early embryos, early embryos, pupae, adult heads and S2 cell libraries. Numbers of reads are normalized to the sequencing deep of each library. The peak of ⬇25nt small RNAs in the head library likely corresponds to piRNAs from contaminating ovaries in the sample (4).

suppresses white and EcR RNAi, in agreement with siRNAmediated target cleavage taking place in the cytoplasm. Development of B2, P19 and NLS-P19 transgenic animals was not altered, nor was the repression of a sensor construct reporting bantam activity (36), indicating no obvious interference with the miRNA pathway by either viral protein (Fig. S1C). Having established that constitutive expression of B2 and P19 specifically impairs hairpin-induced RNAi in living flies, we tested whether they altered the endogenous siRNA pathway. Epitopetagged B2, P19, or NLS-P19 were transiently expressed in S2 cells (Fig. S2 A); 3⬘-end RNA labeling revealed the presence of ⬇21nt RNAs in immunoprecipitates of P19 and, to a lesser extent, of NLS-P19 (Fig. 2A, arrowhead). Among them, siRNAs antisense to

A

19 -P T 9 LS Ladders 2 pM B P1 N 23 27

*

B

I

nt 150

pMT IP

B2 I

P19 IP

I

IP

NLS-P19 I

IP

*

150

30 20

30 20

Fig. 2. P19 and B2 respectively sequester endogenous TE-matching siRNAs or longer precursors in S2 cells. (A) Immunoprecipitated P19 and NLS-P19 sequester ⬇21 nt RNAs that migrate as ⬇22–23 nt species after 3⬘pCp labeling (arrowhead) whereas larger RNA species are sequestered exclusively by B2 (*). Control immunoprecipitation (pMT) was performed using S2 cells transfected with the empty expression vector pMT-DEST48 (B) A sense HMS-Beagle probe revealed enriched siRNAs in Northern blots of P19 and NLS-P19 immunoprecipitates (IP) and longer RNA species (*) in B2 immunoprecipitate. I, corresponds to total RNA input material.

Fagegaltier et al.

the LTR-retrotransposons HMS-Beagle (Fig. 2B) and to transcripts of the roo TE (Fig. S2D) were detected. By contrast, B2 immunoprecipitates contained barely detectable amounts of siRNAs (Fig. 2 A and B). Rather, long (⬎200nt) RNA species were detected (Fig. 2 A, asterisk), including HMS-Beagle and roo antisense transcripts (Fig. 2B and Fig. S2D, asterisk), suggesting that B2 binds long double-stranded TE RNAs and likely compromises small RNA biogenesis. To characterize the small RNAs bound in vivo by P19 and NLS-P19, we performed high-throughput sequencing by Solexa methodology of small RNA libraries from S2 cells expressing P19 or NLS-P19 (P19 and NLS-P19 inputs), as well as from small RNAs libraries coimmunoprecipitated with P19 or NLS-P19, respectively (P19 IP and NLS-P19 IP). We also generated a small RNA library from nontransfected cells (S2) as a control. Deep sequencing of each of the five libraries yielded from 470,136 to 1,050,202 19–29nt RNA reads that matched the D. melanogaster genome. These reads fell into the various RNA classes depicted in Fig. 3A. The proportion of 21nt TE-matching siRNAs ranged from 3.4% to 8% in the libraries from control, P19- and NLS-P19-expressing S2 cells. However, these species were strikingly enriched to 71.3% in the library immunoprecipitated from P19-bound RNAs (Fig. 3 A and B). Although to a lesser extent (17.4%), TE-matching siRNAs were also significantly enriched in immunoprecipitated NLS-P19-bound RNAs. We noted that the proportion of rRNA-matching products increased in parallel in this specific library (Fig. 3A) whereas there was an enrichment in exon-matching small RNAs in P19 and NLS-P19 bound RNAs (Fig. 3A, raw ‘‘exonic’’). Genome mapping of these small RNAs indicated that they mostly correspond to previously described endo-siRNAs produced from cis-natural antisense transcripts (1, 5, 6, 37). In contrast to endo-siRNAs, the proportion of miRNAs found in P19 and NLS-P19 immunoprecipitates decreased dramatically (3.4% and 26.7%, respectively), compared with that found in libraries from control, P19- and NLS-P19-expressing cells (⬇85%) (Fig. 3 A and B). The P19 protein specifically binds 21bp double-stranded siRNAs with 2nt overhangs (27, 28). We thus expected an enrichment of siRNA duplexes showing perfect strand complementarity over 19nt in the libraries from P19 IP and NLS-P19 IP bound RNAs. To test this notion we plotted the distance between each TE-matched siRNA 5⬘ end in the genome and the 5⬘ end of its neighbors on the opposite strand over a window of ⫺100/⫹100nt. We found that the relative probability of finding an siRNA partner whose 5⬘ end can be mapped 19nt away on the complementary strand dramatically increased in the P19 IP and NLSP19 IP libraries compared with the S2, P19 input and NLS-P19 input libraries (Fig. 3C). Likewise, we observed a dramatic increase in the probability of 19bp duplex formation when the analysis was restricted to the siRNAs matching the 297 retrotransposon only (Fig. S3). Altogether, these results demonstrate that P19 and NLS-P19 sequester endo-siRNA duplexes, which correspond in their majority to TE-derived siRNAs. It is noteworthy that both in Northern experiments and high-throughput sequencing, the fraction of endo-siRNAs bound by NLS-P19 appeared reduced as compared with the one bound by P19, which is nearly exclusively cytoplasmic (Fig. S2 A). Nonetheless, the nature of TEs giving rise to the most abundant endo-siRNAs did not change among the 5 libraries, with 297, blood and 1731 retrotransposons invariably accounting for ⬇60% of the TE-matching siRNAs (Fig. S4A). In addition, genomic matches of the TEderived siRNAs did not vary significantly between libraries (Fig. S4B). Therefore, the respective affinity of P19 and NLS-P19 for TE-derived siRNAs does not appear to be biased for particular classes of TEs. Because the two proteins immunoprecipitated at similar levels (Fig. S2C), the results thus suggest the existence of an abundant, cytoplasmic pool of endo-siRNA in S2 cells, in PNAS 兩 December 15, 2009 兩 vol. 106 兩 no. 50 兩 21259

GENETICS

200

A rRNA, tRNA, snoRNA miRNAs TEs exonic intronic intergenic no annotation matched reads in library

P19 input

5,2% 89,9% 3,4% 0,5% 0,3% 0,4% 0,4% 793032

4,6% 83,9% 8,8% 0,8% 0,4% 0,8% 0,6% 767120

NLS-P19 input 5,3% 83,8% 7,9% 0,8% 0,3% 1,1% 0,9% 594778

800 700 600 500 400 300 200 100 0

B

Number of reads in a thousand

S2

P19 IP 9,4% 3,4% 71,3% 5,2% 1,9% 5,1% 3,7% 1050202

NLS-P19 IP 39,4% 26,7% 17,4% 5,8% 2,4% 4,7% 3,6% 470136

S2

19 20 21 22 23 24 25 26 27 28 29 800 700 600 P19 input NLS-P19 input 500 400 300 200 100 0 19 20 21 22 23 24 25 26 27 28 29 19 20 21 22 23 24 25 26 27 28 29 250 600 800 700 600 500 400 300 200 100 0

500

200

P19 IP

400

NLS-P19 IP

150

300 100

200

50

100

0

0 19 20 21 22 23 24 25 26 27 28 29

19 20 21 22 23 24 25 26 27 28 29

small RNA length (nt)

rel. frequency of antisens 5’ end

C 0,6

18

0,3

18

0,5 0,2

0,4 0,3 0,2

P19 IP

NLS-P19 IP 0,1

0,1

NLS-P19 input P19 input 0,0 0,0 -100-80 -60 -40 -20 0 20 40 60 80 100 -100-80 -60 -40 -20 0 20 40 60 80 100 distance from 5’ end (nt)

Fig. 3. Characterization of small RNAs bound by P19 and NLS-P19. (A) Annotation of small RNAs isolated from S2 cells, S2 cells expressing P19 (P19 input) or NLS-P19 (NLS-P19), and P19 (P19 IP) or NLSP19 (NLS-P19 IP) immunoprecipitates. (B) Length profiles of miRNAs (gray bars) and TE-matching small RNAs (black bars) in small RNA libraries. Numbers of reads are normalized to the sequencing deep of each library. (C) Frequency maps, for the P19 input, P19 IP, NLS-P19 input and NLS IP libraries, of the separation of TEmatching siRNAs mapping to opposite genomic strands. The spike at position 18 indicates the position of maximal probability of finding the 5⬘ end of a complementary siRNA, which corresponds to a 19nt offset (graphs start at 0).

addition to a moderately abundant pool of nuclear endosiRNAs. We next tested whether the effects of P19, NLS-P19 and B2 on endo-siRNAs impinged on heterochromatin formation and distribution in adult flies. To this end, we measured their effect on Position Effect Variegation (PEV), a process by which hetero21260 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0809208105

chromatin invasion of a marker gene causes its silencing. The T(2;3)SbV translocation relocates a dominant mutant allele of Stubble (Sb1) from its normal position on chromosome 3R to the 2R pericentromere. The ensuing heterochromatic silencing of SbV results in longer, nearly wild type bristles (38). Ubiquitous expression of B2 and NLS-P19 relieved SbV silencing in adult flies, indicating that B2 and NLS-P19 suppress PEV; in contrast, P19 had no effect (Fig. 4A). When expressed in larval salivary glands, B2 was found in the cytoplasm and only faintly detected in the nucleus or in perinuclear regions; P19 remained cytoplasmic and was enriched at the cytoplasmic membrane whereas NLS-P19 was exclusively nuclear (Fig. S5A). In polytene chromosomes from wild type salivary glands (Fig. 4B), dimethylation of histone H3-K9 residue (H3m2K9) typically covers heterochromatin in the pericentromere, telomeres, and a few loci along chromosome arms. In contrast, ⬇60% of polytene chromosomes from larvae expressing NLS-P19 had poor H3m2K9 labeling at the pericentromere, but showed, in contrast, strong labeling spread across chromosome arms. Chromosomes from B2expressing animals displayed similarly altered H3m2K9 patterns, albeit less frequently (⬇30%). Furthermore, we found that NLS-P19 and B2 strongly increased the pericentromeric distribution of H3m3K9 (Fig. 4B), another heterochromatic mark that normally accumulates at the chromocenter core, and only weakly at the pericentromere in a Su(var)3–9-dependent manner (39). NLS-P19 and B2 also affected the distribution of Heterochromatin Protein 1 (HP1), normally concentrated at the pericentric heterochromatin and the fourth chromosome. Indeed, paralleling the spreading of H3m2K9, strong ectopic HP1 labeling was detected on the arms of NLS-P19 and B2 polytene chromosomes (Fig. 4B), in agreement with a role for the H3m2K9 mark in recruiting HP1 (40). In accordance with the lack of P19 effect on PEV, distribution of H3m2K9, H3m3K9 and HP1 was unaltered in f lies expressing P19 at the same levels as NLS-P19 (Fig. 4B and Fig. S5B). This result indicates that the P19 effect on heterochromatin entails its nuclear localization, as had been previously shown in plants (41). Finally, we tested the distribution of Su(var)3–9, a major and well characterized Drosophila H3-K9 methyltransferase that locates prominently at the pericentromere (39). In animals expressing NLS-P19 in salivary glands, Su(var)3–9 labeling was reduced at the pericentromere and accumulated ectopically along chromosome arms (Fig. 4C), mirroring the unusual H3m2K9 patterns induced by NLS-P19 and B2. Because neither B2 nor NLS-P19 associates directly with chromosomes, these results strongly suggest that sequestering siRNAs or their precursors is sufficient to generate aberrant H3K9 methylation patterns and ectopic HP1 localization on chromosomes. Collectively, the data suggests that endo-siRNAs are required for heterochromatin silencing in the adult soma and for proper targeting of Su(var)3–9 and H3K9 methylation at the pericentromere. To test this model further, we analyzed the effect of mutations in RNAi pathway components on PEV. These mutations are expected to alter the biogenesis or activity of endo-siRNAs (1, 2). The Su(var)3-9 mutation, used as a reference, eliminated SbV silencing in the T(2;3)SbV test strain. This silencing was also significantly compromised in heterozygous dcr2, r2d2, and ago2 mutants but not in loqs mutant (Fig. 5A). Because RNAi mutations are exclusively of paternal origin in this experiment, the result indicates that Dcr2, Ago2, and R2d2 are zygotically required for SbV repression. Notably, the inhibiting effect of the dcr2G31R mutation, which specifically inactivates the nuclease function of Dcr2 while presumably keeping its dsRNA binding property intact (42), suggests that processing of siRNAs, is mandatory for SbV silencing. We carried out similar analyses on the white-mottled 4 rearrangeFagegaltier et al.

A 110 p=0.23 100 90

2,7E-06

120

7,9E-12 3,4E-9

ns

100

op 1

4

ag o2 d

O

ag o2 41

lo qs K

d2

R

r2

31

6X

r2 G dc

41

sX

dc

(v su

H3m2K9

r2 R

+

P19

1

80

NLS-P19

dc

B2

NLS-P19

2,0E-23 3,9E-07 2,8E-16

140

.9 6

P19

160

ar

+

B2

1,2E-70

180

81 1f

+

200

)3

80

B

pH2b-YFP

lio-GAL4>NLS-P19

D

lio-GAL4>H2b-YFP

Ago2414/Ago2414

F

lio-GAL4>H2b-YFP

lio-GAL4>P19

lio-GAL4>H2b-YFP

Dcr2R416X/Dcr2R416X

HP1 HP1 H2b-YFP

H2b-YFP

H2b-YFP HP1 H2b-YFP

E

lio-GAL4>B2

H2b-YFP

H2b-YFP HP1

C

lio-GAL4>H2b-YFP

HP1

lio-GAL4>-

HP1

A

Fig. S8. B2, NLS-P19, ago2 and dcr2 mutations affect HP1 distribution on polytene chromosomes from third instar salivary glands. (A) Salivary glands from a lio⬎GAL4/⫹ and from a lio⬎GAL4/⫹; UAS⬎H2b-YFP/⫹ larvae were squashed on the same slide and stained with DAPI (blue) and anti-HP1 antibody (red). Chromosome sets were genotyped owing to yellow fluorescence of YFP. H2b-YFP expression does not affect HP1 distribution. Control salivary glands from lio⬎GAL4/⫹; UAS⬎H2b-YFP/⫹ (Left) and salivary glands from lio⬎GAL4/⫹; UAS⬎B2/⫹ (B), lio-GAL4/⫹; UAS⬎P19/⫹ (C), lio-GAL4/⫹; UAS⬎NLS-P19/⫹ (D), ago2414/ago2414 (E) or dcr2R418X/ dcr2R418X (F) larvae (Right) were spread on the same slide. Chromosome sets were stained and genotyped following the same procedure.

Fagegaltier et al. www.pnas.org/cgi/content/short/0809208105

10 of 10